There might be good reasons for Britain to build data centres. Job creation isn’t one of them
Estimate*estimate
governments put out unambiguous figures: the amount you owe in tax, the number of hospitals we need to take care of sick children, the price of small caged mammals. Sometimes governments put out numbers that are based on estimates.
The UK government has made a lot of noise about an “AI growth zone” with a massive data centre complex on the site of the shuttered Britishvolt facility near Cambois, Northumberland. They said those server huts, developed by the Blackstone-owned QTS, would create 4,000 jobs.
Here’s Sarah O’Connor in MainFT last year, on that eyebrow raising jobs promise:
According to the planning documents, the development “is estimated to require up to 40 people per data centre”. Once all 10 centres are built (expected by 2035) that makes 400 jobs, a tenth of the number the government cited.The planning documents say “peak monthly construction employment could be in the region of 1,200 jobs” — these aren’t permanent roles but there is likely to be a decade’s worth of construction work on the site, so they aren’t fleeting jobs either. Even so, the operational and construction jobs combined only account for 1,600 jobs. Where are the rest?[ . . . ] The planning documents for the Northumberland project say that one direct job in a data centre can support between five and seven additional jobs in the local economy and related industries, which “could equate to an additional 2,000 to 2,800 jobs in the local economy once all ten data centre buildings are operational”. Take 2,400 as the midpoint and we have reached the government’s 4,000 number.
Where are these numbers coming from? The estimate the government cited is actually from Northumberland County Council.
And how did they come up with that estimate, you ask? See this Freedom of Information request response, shared with Alphaville by the tech campaign group Foxglove:
So QTS, the company trying to convince the council to let them build the data centre, were the original source of the jobs estimate. Off to a good start.
Since then, two things have happened.
The first is that the government put out a new press release declaring a ‘North East AI Growth Zone’, which essentially just bundled together that 4,000 jobs Cambois data centre and another nearby data centre, Cobalt Park, which has another very prominent tenant: OpenAI. Cobalt Park will host a data centre for OpenAI’s totally unprofitable Stargate project. The total for that North East AI Growth Zone is now listed at 5,000 jobs (again, this is mainly made up of the previously announced Cambois data centre)
What makes all of this even more interesting is the second new thing, an FOI that Foxglove put forward, which asked about the jobs estimate for another AI Growth Zone, this one in faraway South Wales, which the government says will create “more than 5,000 jobs”. Microsoft has already signed up, and released a 3D render that will really get the blood pumping:
Please use the sharing tools found via the share button at the top or side of articles. Copying articles to share with others is a breach of FT.com T&Cs and Copyright Policy. Email licensing@ft.com to buy additional rights. Subscribers may share up to 10 or 20 articles per month using the gift article service. More information can be found at https://www.ft.com/tour.
So how did the government reach that figure of 5,000 jobs for the handsome new South Wales project? Again, from that FOI:
“The “more than 5,000 jobs” estimate is based on conservatively scaling projected jobs using evidence from the Cambois data centre development [. . .] and adjusting these figures for the significantly higher capacity of the South Wales AI Growth Zone.
How much “significantly higher” is the capacity, you ask? In the FOI response, the government says it’ll be big. So big, in fact, that they had to search far and wide for a source to back up their estimates:
Leaving aside the use of generative AI to estimate the jobs created by generative AI, the FOI says the government applied a 1.5-1.6 jobs multiplier on construction and operational jobs. That’s not an unreasonable assumption for the construction industry, but when it comes to data centre employment, things aren’t so clear cut.
The government gives what’s probably the most relevant source for their estimate via a report authored by the industry lobby group Tech UK. Per the FOI:
TechUK suggests that a typical 5.7-7MW UK data centre will employ between 20 and 88 roles in full time equivalent (FTE) operational employment. This does not account for indirect and induced job creation mentioned above. The total South Wales AI Growth Zone development is anticipated to reach hundreds of megawatts of capacity, therefore likely creating thousands of direct, indirect and induced roles across the life cycle of the project.”
The assumption here is that more mega-wattage means more jobs. The government links the TechUK report appendix as a source:
Alphaville asked Tim Anker, who runs the Colo-X brokerage and consultancy cited by Tech UK for the top end of that megawatt average. Anker says he was surprised to see that the government was referencing the database of data centres hosted publicly on his website. Firstly, because the total UK mega-wattage the government cites is from 2024 and therefore now well out-of-date (he estimates the total number now is closer to 1,500MW).
Secondly, and much more importantly, because there’s a deeper issue with all of the averages cited by the government: the difference between the much more common “co-location” data centres and the big “hyperscale” data centres run by the likes of Microsoft or OpenAI.
Colo-X specifically tracks co-location sites, which have much smaller average energy usage and multiple customers using the same building. Which crucially means a higher number of employees are needed.
Simply dividing the total energy use by the number of data centres ignores the fact that the economics of these kinds of co-located data centres — some of which were built twenty or thirty years ago — are, as the name might suggest, completely different from hyperscale sites.
“What was built back then to service the needs of lots of small customers in no way reflects a current building boom,” Anker told FTAV. “The staffing of a small co-location data center with lots of small customers will be quite high. Whereas when you’ve got one large hyperscale, ultra efficient user, staffing levels will be much lower.”
Or to put it more bluntly:
They’re definitely going to overestimate the number of jobs created by assuming that hyperscale megawatts have the same employment ratio as retail co-location facilities, which they don’t.
On the jobs front, the top of TechUK’s range, 88 jobs, is an estimate provided by Equinix, who are — surprise, surprise — a co-location data centre provider. The report they cite even has a content warning: “Equinix’s digital infrastructure is different to the hyperscale / cloud data centre”.
But even the most customer-focused co-location data centres, which can offer a variety of web-hosting services (rather than raw compute power), aren’t huge job creators, Anker says. “Once it’s up and running the on-site staff is a technical support and security, basically.”
On the government’s attempt to extrapolate to the larger rural sites, Anker pointed out that the hyperscalers aren’t particularly inclined to pack these facilities out with hardworking employees.
“They’re renowned for being ruthlessly efficient,” he told us. “I find it baffling that the government is so enthralled with these American hyperscalers.”
Ultimately, the economies of scale involved means larger sites create less jobs per megawatt than the smaller, regional co-location sites used for those estimates.
So it turns out trying to project up data centre job figures is tricky. Per Sarah O’Connor again:
There are plenty of good arguments for the economic importance of data centres (as well as some concerns about energy and water use), but “they create tonnes of local jobs” isn’t one of them.
Could the government offer its support to some more widely beneficial real-estate ventures? While we’re having fun using megawatt averages and “scaling projected jobs”, Alphaville has selected a few more exciting examples to match those 20-88 full time roles:
You could build a mere 22 per cent of a sparkling new leisure venue with an indoor go-karting track in Newcastle city centre (expected to create 400 jobs)